Despite the misconception that automation tools like ChatGPT reduce job opportunities, the World Economic Forum (WEF) estimates that 97 million jobs will be created by 2025 as a result of AI. This means that there are plenty of opportunities to upskill the workforce to keep pace with automation.
As IT talent shortages remain an ongoing challenge in 2023, tech leaders are forced to implement new strategies to keep their organizations equipped for the digital future. The demand for tech talent continues to grow, how are leaders responding to the skills gap? Are forward-thinking tech executives reskilling & upskilling their workforce, using AI to augment human capabilities, or investing in a combination of both?
With this as background, “Tech in Motion” brought together several AI strategic leaders to discuss the issues and trends: Moderator David Yakobovitch Global Product Lead @ Google; and panelists Howie Xu Vice President of Machine Learning and AI @ Zscaler; Jennifer Glenski Director of Product Management @ BMC Software; Greg Coquillo AI, Technology Manager @ Amazon and Daniel Wu Head of AI & Machine Learning, Commercial Banking JPMorgan Chase & Co.
This panel of experts shared their advice for using automation to lead digital transformation while continuing to nurture talent and empower employees. To be the first to rewatch the event once it goes live on-demand, sign up at the event's home page: Automation, Reskilling, and the Changing World of Work.
Note: The opinions expressed are the panelist's personal thoughts and are not representative of the organizations they work for.
How do skills change with new automation priorities in 2023? How are you addressing skill gaps?
Coquillo: If you don’t have in-house expertise, look outside and try to hire the right people. Most of the time large organizations do have competent employees. They just need to fill gaps with diverse skill sets. Fortunately, today there is a large pool to draw from.
There is also a need to put in programs to upskill existing talent. It’s important to note that not everybody needs to have a full master’s course in Computer Science. Different people have different needs. Off-the-shelf or in-house solutions are options. Collaborate across the board within your organization, across different lines of business to obtain common skills. Design your programs accordingly, and don’t err on side of spending a lot of money on training which might be overkill.
Xu: A big change for us is a mindset and culture shift as a team and as individuals. We are at a tipping point with AI. Maybe not human intelligence yet, but very powerful, nevertheless. How do we adapt and cope with it? How do we replace tedious work? I see this as big as the internet wave. We should embrace it. Get our mindset ready. The world is going to be very different, and organizations will need to be bold with their restructuring.
How are you upskilling within your own job?
Coquillo: I am a sucker for acquiring new skills. I focus on the need to change my skillset first, then find resources. Lately, I have been creating mental models for new challenges. How do you create mental models on demand for situations you have never seen before, and find people to help reinforce your mental model? What tools, skills and resources do you need to address them?
Wu: I am always learning. One big area I have focused on is Responsible AI. What does responsible AI mean? What are the principles that are practiced?
Secondly, is on the leadership side. I am focusing on setting strategies and visions for leading technologies that enable transformation. How to put in place strategies to move a large organization to use AI technologies is a large task. There are no recipes to follow. You have to learn about what other industries are doing to drive transformation.
At a technical level, AI requires a data-driven approach. Collecting, analyzing, and getting insights out of data. This is not the typical way of doing things in the last couple of decades. What does getting insights from data mean? We have to learn from that, both on the technical and business side.
Glenski: I ask myself “what are you upskilling”? In my role in product management in innovation labs, I try to anticipate customer problems of tomorrow. I enjoy self-paced upskilling and reskilling in my own time without going back to school full-time, which I do not have time for. On the technical side, I am interested in working on applying AI to the data ops world, similar to IT operations.
.png?width=900&height=225&name=Copy%20of%20Blog%20Call%20Outs%20(21).png)
What about the talent needed to build application and infrastructure layers, or AI algorithm applications?
Coquillo: In order to deploy a portfolio of AI products to be successful, we need data scientists, machine learning engineers, software engineers and data engineers. One thing people don’t talk enough about is supporting teams to make machine learning models work. In the room, you need an AI architect to take stock of AI models deployed to product teams to assess the ROI. Legal to cover all ethical and privacy issues around AI. Finance and accounting to make sure generated revenues are going to the right accounts. Program teams to support workflows going around. They are critical to understanding where AI can help, and where they need to step back.
Yakobovitch: Yes, both the business roles as well as technical roles. Different divisions such as service management, IT operations, data operations, data management and security operations need to sharpen their focus as well.
Glenski: Some of the IT processes we look at are the same because these environments are always evolving and sometimes it can be difficult to keep up by automating them to handle mundane repetitive processes. One of the benefits of automation over manual is to jump-start AI operations to take the burden off people and put it in on machines. This frees up employees to work on innovation and higher-value projects.
Xu: One of the latest breakthroughs of AI is reinforcing human learning with AI in the loop. On the human side, how do we collaborate, and at what place?
What’s top of mind when it comes to automation trends?
Xu: If you are not interacting with ChatGPT on weekly basis you are not doing enough. If you work in marketing, programming, legal, etc. you should you digging through it. What is the limit? Each one of us in an enterprise needs to understand what it means to us. It will be a different user experience for each of us.
Wu: I still want to ground the discussion on the societal economic impact of introducing ChatGPT to our workplaces. There are lots of loose ends still, with all this excitement. How does AI change education for example? Some schools are banning ChatGPT because teachers cannot assign essays without knowing whether students are using ChatGPT to cheat.
Nobody is doing a holistic evaluation as far as responsible AI metrics. This is a key area to increase and build trust in AI. Lots of people are still on the fence and don’t want to dive right in. They need to know this is here to stay and sustainable, and there are mitigation strategies in place.
Glenski: One trend I am excited about is Automation with AI ops. For example, a user submits a question and gets a resolution and direction from AI. Typically, an IT admin gets an alert, finds an error code, and googles it to get a resolution – all this could be done away with ChatGPT, which also knows the vulnerabilities and patches which are out there to address the root cause.
Coquillo: I am excited about the birth of more large language models with big vision models. What about a robotics model that has physical machines embedded with AI models to augment people’s lives? To train an assistant in the kitchen or something else in the house, such as cleaning up.
Yakobovitch: Some models show that AI can achieve human-level parity with 80-90 percent accuracy.
The number one question from our audience: What category of jobs will AI take over first?
Xu: It’s controversial. For years, the conventional wisdom has been that it will be the repetitive mundane jobs first, but not creative jobs, or senior-level jobs. However, AI can do things very well, like writing a speech for a CEO, and augmenting their job, but not replacing it.
Yakobovitch: Yes, cold start problems. It takes too much time to start writing or to start coding, so if we can have this augmentation then humans can focus on more complex, riveting work.
Coquillo: The cold start problem is because the human brain is chaotic and often doesn’t know where to start in addressing a task or problem. ChatGPT can make sense out of that chaos, replacing the cognitive overload that we don’t have the bandwidth or energy for.
What are some of the risks with large language models?
Wu: One thing that is common across Generative AI, is hallucinations. A computer hallucination is an artificial intelligence (AI) machine vision and machine learning technology interpretation error. Computer hallucinations are created by a variety of factors that cause AI systems to misclassify. Yes, it learns from large data sent but it is unreliably confident about the results.
Another risk is associated with bias. Humans are inherently biased at individual levels. As we automate and train these large data models, we are propagating this collective bias. It can cascade and spread like a wildfire. We need to understand and mitigate these biases. It is not just a nice to have but a legal requirement. First, we have to define what is bias and how to measure it. This is no easy task.
